Predictive modelling of drilling temperature in decorative particleboard via machine learning

Drilling is the primary method for machining holes in decorative particleboard. The amount of heat generated in this process is a key factor affecting drilling performance. In this work, drilling temperature was investigated using response surface methodology (RSM) and backpropagation neural network (BP NN) at different machining parameter combinations of spindle speed, axis feed rate, and point angle of tool during drilling of melamine-impregnated paper decorative particleboard. Results showed that increased spindle speed and feed rate led to higher drilling temperatures, while drilling temperature was positively corelated to the point angle of the tool. Based on both variance analysis and sensitivity analysis, the feed rate is identified as the most influential parameter on drilling temperature, followed by spindle speed and point angle; all three factors contributed significantly to drilling temperature. Furthermore, prediction models developed through RSM and BP-NN all had high prediction accuracy with errors of 4.83% and 3.05%, respectively. However, due to its lower error, BP NN resulted in more effective optimisation compared to traditional RSM. Finally, optimal drilling parameters, based on lowest temperature, were determined for melamine-impregnated paper decorative particleboard to be spindle speed of 4,007 rpm, feed rate of 1,974mm/min, and point angle of 60°.